Purpose: Instrumental voice assessment plays a critical role in identifying vocal issues and for documenting treatment outcomes. The reported voice data, however, are sensitive to the algorithm used by each acoustic analysis software program (AASP) to analyze the corresponding waveform. In the present study, five acoustic measures were compared across healthy speakers and speakers with dysphonia for three AASPs commonly used in research, education, and clinical practice: Multidimensional Voice Program (MDVP) by Computerized Speech Lab, Praat, and TF32.
Materials and methods: Sustained vowel phonations for the quantal vowels /ɑ/, /i/, and /u/ were analyzed for 80 speakers with organic dysphonia and 60 age- and sex-matched healthy controls. Descriptive, inferential, and correlation data are reported for mean fundamental frequency (mean F0), standard deviation of fundamental frequency (SD F0), short-term perturbation measures of jitter and shimmer, and harmonic-to-noise ratio (HNR).
Results: The present study replicated previous findings of interprogram differences for healthy speakers, with MDVP consistently yielding higher values than Praat and TF32 for SD F0, jitter, and shimmer and lower values for HNR. Similar, but magnified patterns of results were observed for speakers with dysphonia.
Conclusion: The variation observed across programs calls into question the validity in comparing voice outcomes reported by one AASP to those previously obtained by another, particularly for acoustic signals with aperiodic components that are commonly present in disordered voices. It is advised that waveforms be visually inspected prior to conducting acoustic analysis, and that voice outcomes not be combined or compared across AASPs.
Keywords: Acoustic voice analysis; Vocal perturbation; Voice disorders.
Copyright © 2022 The Voice Foundation. Published by Elsevier Inc. All rights reserved.